INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Mouse mammary tumor virus (MMTV) is a retrovirus that is either inherited as an endogenous virus or acquired as an exogenous virus through milk-borne infection. MMTV has been used as a model for the study of the regulation of gene transcription since the discovery that its expression was induced by glucocorticoid hormones in vivo and in tissue culture cells (reviewed in reference 61). Indeed, the first evidence that mammalian transcription factors interacted with specific DNA sequences (termed glucocorticoid response elements [GREs]) came from studies of how glucocorticoid receptors (GR) induced MMTV expression (61). The ability of glucocorticoids and progesterone to stimulate viral transcription is critical for MMTV transmission to subsequent generations, since as a result of this stimulation, virus production dramatically increases during pregnancy and lactation (5).

A number of additional transcription factors, including NF-1, Oct-1, and TFIID, are involved in the regulation of MMTV gene expression (10, 40, 54). Moreover, as expected for a virus transmitted through milk, there are sequences within the long terminal repeat (LTR) of the virus that confer mammary gland-specific expression, termed the mammary gland enhancer (9, 27, 40, 41, 47) (Fig. 1A). Transgenic mouse studies in which this enhancer was linked to the heterologous simian virus 40 promoter indicated that it directed expression to lactating and virgin mammary gland that was no longer lactation responsive (41). Inclusion of the GREs in the transgene restored lactation-induced expression. Several transcription factors, including AP-2 (56) and NF-1 or related factors (27, 40), have been shown to bind to this region.

View larger version (21K): [in this window] [in a new window]   FIG. 1.   (A)Map of the MMTV LTR. Depicted are the binding sites for Oct-1 and NF-1 transcription factors, as well as the GREs. MGE refers to the region mapped as a mammary gland enhancer at the 5' end of the LTR. The region with homology to STAT binding sites (bp 519 and 528) is also shown. HR denotes the coding region for the hypervariable domain of the Sag. (B) LTRs of the constructs used to create transgenic mice. The filled box represents the MMTV(C3H) LTR; the open box represents the Mtv-7 LTR. The Mtv7/C3H has the regions from bp 1 to 631 from the Mtv-7 LTR and from bp 632 to 1280 from the MMTV(C3H) LTR, and the C3H/Mtv7 LTR has the regions from bp 1 to 631 from the MMTV(C3H) LTR and from 632 to 1280 from the Mtv-7 LTR. The wide stripes represent the STAT region from MMTV(C3H); the narrow stripes represent the STAT region from Mtv-7. (C) Pictorial representation of the Mtv-7/BALB14 naturally occurring recombinant LTRs. Two classes of recombinants are shown, those that had a recombination event between the HR and STAT binding sites of BALB14 and Mtv-7 (REC 2; the actual breakpoint differed from virus to virus) and those with a breakpoint within the STAT sequences (REC 1). Both types of events result in a virus with the hypervariable region from Mtv-7 and the T · A base pair at position 520 in the STAT site.

In addition to mammary gland cells, lymphoid cells transcribe MMTV (9, 17, 22, 26) and shed virus particles (12). MMTV expression in these cells is critical to the virus life cycle, since infected B cells in newborn pups present an MMTV-encoded superantigen (Sag) to cognate T cells (3); this activation of T lymphocytes by Sag is requisite step in the virus's ultimate transmission to the mammary gland (14). Sag proteins interact with all T cells that express particular  chains of the T-cell receptor (38). This interaction is determined by the sequence of the C-terminal amino acids of this type II membrane glycoprotein and different MMTVs associate with different V-bearing T cells, depending on the sequence of this hypervariable region (1, 4, 8, 44). Little is known about what positively determines expression of the virus in lymphoid cells and where the regulatory regions controlling this expression lie, although it has recently been shown that a nuclear matrix bound transcription factor called SATB1 negatively regulates MMTV transcription in lymphocytes (31).

Recently, several milk protein genes, such as those encoding the whey acidic protein (WAP), -casein, and -lactoglobulin proteins, were shown to have a common transcription-regulatory element that bound a factor in lactating mammary gland (7, 29, 50). This factor was purified and shown to be a member of the STAT family of transcription factors (58), and it was subsequently called STAT5. STAT5 is activated in response to prolactin through phosphorylation by the Janus (JAK) family of tyrosine kinases (24, 49, 59). There are two STAT5 genes in mice, called Stat5a and Stat5b, that presumably arose through gene duplication. It is unclear whether their function is duplicated as well (32). Although STAT5 was thought to be important for the tissue-specific expression of milk protein genes, mutation of the STAT consensus binding sites in the WAP or -lactoglobulin gene promoter had only subtle effects on linked marker gene expression in lactating transgenic mice (7, 30). However, in mice lacking Stat5a, WAP expression was down-regulated, but little or no effect was seen on -casein, -lactoglobulin, or WDNM1 (a milk protein) RNA levels (33). Similarly, it has been reported that in Stat5b knockout mice, there is transcription of milk protein genes (56). It is now clear that Stat5a and Stat5b are expressed in wide variety of tissues and are activated by a number of different cytokine or growth factor receptors (24, 49, 59).

We recently found that while most endogenous MMTVs are expressed in both mammary gland and lymphoid tissue (ML expression), others are expressed only in lymphoid cells (L expression) (17, 46). Because this was true of proviruses at various chromosomal locations, it was unlikely that this tissue-specific expression was the result of position effects and was more likely due to sequences within the viruses. We also discovered that a lymphoid cell-specific endogenous MMTV, Mtv-7, could be converted to a mammary gland-expressing exogenous virus through retroviral recombination with a mammary gland-expressing exogenous virus (16).

We show here that there is a single base pair difference in a transcription factor binding site in the LTR between the lymphoid-specific MMTVs and those expressed both in mammary gland and lymphocytes. All MMTVs could be grouped according to this consensus sequence into viruses exhibiting ML expression and those exhibiting L expression. Moreover, there was preferential binding of a complex isolated from mammary cell nuclear extracts to the ML sequence that was inducible by prolactin and was supershifted by anti-Stat5b antibody. In mice that lacked the Stat5b or Stat5a transcription factor, transcription of endogenous MMTVs was dramatically reduced in mammary gland but not lymphoid tissue. These results indicate that there is a site in the MMTV LTR that dominantly affects mammary gland transcription of the virus and that STAT5 or a related factor plays a role in the control of MMTV's tissue-specific expression in vivo.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Cell culture. The NMuMG (normal murine mammary gland) cell line was obtained through the American Type Culture Collection and grown according to the supplier's instructions. For prolactin induction, the cells were grown to confluence, and prolactin (5 µg/ml; Sigma Chemical Co., St. Louis, Mo.) or dexamethasone (0.5 µM) was added to the medium. The cells were harvested at 14 to 16 h poststimulation. The NMgCl1 cell line was derived by cotransfection of NMuMG cells with a molecular clone of MMTV called HYB PRO (51) and pSV2neo and then selected in G418 (400 µg/ml; GIBCO/BRL, Gaithersburg, Md.) as previously described (12).

Transgenic mice. Swiss Webster mice (males and females) purchased from the National Institutes of Health Frederick Cancer Research Facility, Frederick, Md., or B6SJLF1/J mice from The Jackson Laboratory were used to make transgenic mice. DNA was microinjected, and transgenic mice were identified by PCR, using primers specific for the simian virus 40 small-t-antigen splice site (62) (not shown). Southern blot analysis of tail DNA was carried out with a probe specific for the chloramphenicol acetyltransferase (CAT) transgene. Quantitation was performed with a PhosphorImager (Molecular Dynamics, Inc., Sunnyvale, Calif.). The LTR-CAT mice, containing the MMTV(C3H) LTR, were previously described (47).

Transgene construction and CAT assays. The Mtv-7 LTR was subcloned from plasmid pMO-BC (4), a kind gift from Brigitte Huber, and inserted into the pCAT^basic plasmid (Promega Biotech, Inc., Madison, Wis.) to make the Mtv-7 transgene. The MMTV(C3H) LTR was derived from the pLTR plasmid (36). To create the Mtv7/C3H and C3H/Mtv7 transgenes, the LTR-containing plasmids were restricted with Eco0109I, which releases a fragment containing the region from bp 632 to 1280 in both the Mtv-7 and C3H LTRs (numbering is according to reference 6); the equivalent Eco0109I fragment from the different LTR was inserted in its place (Fig. 1).

RNA analysis. RNA was isolated by using QuickPrep kits (Pharmacia Biotech, Inc., Uppsala, Sweden) and then subjected to Northern blot analysis and hybridization with an MMTV envelope gene probe. The blot was subsequently stripped and hybridized to a mouse -actin probe, to control for RNA integrity. Quantitation was performed with a PhosphorImager (Molecular Dynamics). The intensity of the hybridization signal of the MMTV bands was normalized to the -actin signal.

Nuclear extract preparation. The cells were harvested and pelleted by centrifugation for 5 min in a tabletop centrifuge at 1,000 rpm at room temperature. Nuclear extracts were prepared as previously described (11, 60). All steps were carried out at 4°C. Phosphatase inhibitors (1 mM sodium fluoride and 100 µM sodium orthovanadate) and protease inhibitors (leupeptin, pepstain A, and aprotinin [5 µg of each per ml] and 0.5 mM phenylmethylsulfonyl fluoride) were included in the buffers at all steps. Protein concentration was measured by the colorimetric method (Bio-Rad Inc., San Rafael, Calif.).

Preparation of oligonucleotide probes. [-³²P]ATP end-labeled oligonucleotide probes were used for gel mobility shift assays, supershift assays, and competition experiments. The double-stranded oligonucleotides were end labeled with 150 µCi of [-³²P]ATP (6,000 Ci/mmol; NEN) in the presence of 10 U of T4 polynucleotide kinase (New England Biolabs, Beverly, Mass.) at 37°C for 1 h. The labeled oligonucleotides were purified by passage through ProbeQuant G₅₀ microcolumns (Pharmacia Biotech). The sequences of the oligonucleotides used for [-³²P]ATP labeling (upper strands only) are as follows: -casein probe, 5'-GGACTTCTTGGAATTAAGGGA-3' (59); FcR1 probe, 5'-GTATTTCCCAGAAAAGGAAC-3' (23); C3H probe, 5'-CTCAACCTCAATTGAAGAACAGTT-3'; Mtv-7 probe, 5'-CTCAACCGCAGTCAAAGAACAGTT-3'; Mtv-43 probe, 5'-CTCAACCTCAGTCAAAGAACAGTT-3'; NFB probe, 5'-GATCTTTGGCTTGAAGCCAATA-3' (60).

Gel mobility shift, supershift, and competition experiments. For binding studies, the reaction mixtures contained 10 µg of protein extract, 10 µg of poly(dI-dC) (Sigma), and 0.5 ng of ³²P-end-labeled oligonucleotide probe (2 × 10⁴ cpm) in D₂ buffer (20 mM HEPES [pH 7.9], 0.25 M sucrose, 0.2 mM EDTA, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride) in a total reaction volume of 20 µl. Reaction mixtures were incubated on ice for 30 min and loaded onto 5% polyacrylamide gels containing 22.5 mM Tris-HCl, 22.5 mM borate, and 0.5 mM EDTA). Electrophoresis was for 2 h at 20 mA at room temperature. After electrophoresis, the gel was transferred to 3MM paper, dried, and exposed to Kodak XAR-5 film overnight at 70°C. For the supershift assays, 10 µg of anti-STAT antibodies (Santa Cruz Biotechnology, Inc., Santa Cruz, Calif.) were used. The following polyclonal rabbit anti-mouse antibodies were tested: polyclonal anti-Stat5a (anti-carboxy terminus [SC1081]), anti-Stat5b (anti-carboxy terminus of the p80 protein [SC835]), anti-STAT4 (anti-amino terminus [SC485]), and anti-STAT6 (anti-carboxy terminus [SC981]). Antibodies were incubated with the nuclear extract for 30 min at room temperature prior to addition of the oligonucleotide probe. Reaction mixtures were loaded on a 5% polyacrylamide gel and run at room temperature as described above. For the competition assays, nuclear extracts were incubated with the specified amounts of unlabeled oligonucleotide competitors and the labeled oligonucleotide probe for 30 min on ice. To determine relative affinities, 100- and 800-fold molar excesses of the unlabeled oligonucleotide competitors were used to inhibit the binding of a constant amount of nuclear extract to the probes.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Comparison of sequences of different MMTVs. Endogenous MMTVs can be divided into two groups, ML and L, based on their expression in mammary gland (17, 46). For example, Mtv-7 and -9 are expressed only in lymphoid tissue, while the exogenous viruses MMTV(C3H) and MMTV(SW), as well as several endogenous MMTVs (Mtv-1, -3, -6, and -43) are expressed in both lymphoid and mammary gland cells (Table 1). The LTRs of the different endogenous MMTVs are remarkably homologous to each other outside the superantigen hypervariable coding region (6). There is only one additional region of dissimilarity among the different proviruses, between bp 519 and 528. As shown in Table 1, there are several nucleotide changes in this region. For example, all of the viruses shown except MMTV(C3H) have a G · C base pair at position 523, and some viruses [MMTV(SW), Mtv-7, and Mtv-43) have CA · GT rather than a TG · AC at positions 525 to 526. Significantly, all MMTVs that are expressed only in lymphoid tissue (i.e., Mtv-7 and -9) have a T·AG·C change at position 520. Although MMTV(SW), Mtv-7, and Mtv-43 all interact with V6-bearing T cells and have almost identical Sag hypervariable regions (6), only MMTV(SW) and Mtv-43 are expressed in both mammary gland and lymphocytes (46). This finding indicated that the presence of a T · A base pair at position 520 was important for mammary gland expression.

Tissue-specific expression is determined by the LTR. We and others have previously reported that the LTR from the MMTV(C3H) milk-borne virus directs expression of linked transgenes to both mammary gland and lymphoid tissues (9, 22, 52). To further show that the sequences within the region from nucleotides 518 to 528 were required for mammary gland-specific expression, we constructed three types of transgenic mice. One set of mice had the Mtv-7 LTR (Mtv-7; four independent lines), the second had a hybrid LTR (Mtv7/C3H) with sequences upstream of bp 632 from Mtv-7 and sequences downstream of this site from the MMTV(C3H) LTR (Fig. 1B, Mtv7/C3H) (one line), and the third had a hybrid LTR (C3H/Mtv7) with sequences upstream of bp 632 from MMTV(C3H) and downstream from Mtv-7 (Fig. 1B, C3H/Mtv7) (three lines). All three LTRs were inserted in the same position in a CAT expression vector. The mice were sacrificed, and CAT assays were performed with extracts prepared from various tissues. In addition, the DNA of these mice was examined by semiquantitative Southern blot analysis, to obtain an approximate estimate of the transgene copy number present in the offspring derived from different founder animals (not shown).

View larger version (16K): [in this window] [in a new window]   FIG. 2.   CAT activities in Mtv-7 and Mtv7/C3H transgenic mice (A), in tissues from C3H/Mtv7 and MMTV(C3H) LTR-CAT transgenic mice (data for the latter are taken from reference 47) (B), and in a high-copy-number Mtv-7 transgenic mouse (C). Abbreviations: LIV, liver; LN, lymph node; THY, thymus; SP, spleen; SG, salivary gland; LMG, lactating mammary gland.

STAT consensus sequence binding. The results presented above provided strong biological evidence that the T · AG · C change in the lymphotropic LTRs was responsible for the lack of expression of these viruses in mammary gland. Comparison of these sequences between gp 519 and 528 to those recognized by known transcription factors revealed that they were similar, although not identical, to those bound by STAT family members (Table 1). If the STAT-like sequence in the LTR was a recognition site for a transcription factor present in mammary gland, it was possible that the nucleotide change in the Mtv-7 LTR prevented binding and thus expression in this tissue.

View larger version (55K): [in this window] [in a new window]   FIG. 3.   Analysis of STAT-like DNA-protein complexes in NMuMG cells by EMSA. NMuMG nuclear extracts (10 µg) were incubated with the indicated 32P-labeled probes. The free probe was run off the gel.

View larger version (54K): [in this window] [in a new window]   FIG. 4.   Oligonucleotide competition. The prolactin-induced DNA-protein complexes were analyzed by using 32P-labeled C3H and Mtv-7 probes. The unlabeled oligonucleotide (Oligo) competitors indicated above the lanes were added as indicated. The competed DNA-protein complexes are indicated by the arrows.

Stat5b is found in a prolactin-inducible complex on the ML sequence. Prolactin induces transcription of several milk protein genes in lactating mammary gland, presumably through the phosphorylation and activation of the DNA binding activity of STAT5 (58). Transcription of MMTV also increases during lactation, when prolactin levels are elevated; however, much, if not all, of this induction is attributable to glucocorticoid hormone induction (41). Although the STAT consensus sequences in the MMTV LTR differed somewhat from those found in milk protein and other genes that have previously been shown to be transcriptionally regulated by STAT5 and other members of this family (Fig. 1), it was possible that prolactin would induce the activity of the factors that bound to the ML- or L-STAT sequence. To test this, nuclear extracts were prepared from NMuMG cells that were treated with prolactin and used in EMSAs within the MMTV(C3H) ML-STAT or Mtv-7 L-STAT probe. Prolactin treatment resulted in induction of the more slowly migrating complex that formed on the ML-STAT (Fig. 5, lanes 2 and 4) but not the L-STAT (lane 8) probe. This was further evidence that the factor(s) recognizing the ML-STAT sequence differed from that bound to the L-STAT sequence.

View larger version (63K): [in this window] [in a new window]   FIG. 5.   Prolactin induction and supershift analysis with anti-STAT5 antibodies. Nuclear extracts were prepared from both uninduced and prolactin-induced NMuMG cells. Nuclear extracts (10 µg) from prolactin-induced cells were incubated with 10 µg of polyclonal anti-Stat5a or anti-Stat5b antibodies at room temperature for 30 min prior to addition of 32P-labeled C3H (lanes 1 to 6) and Mtv-7 (lanes 7 to 10) probes. Arrows indicate the two complexes that form on the C3H probe as well as the supershifted DNA-protein complex formed in the presence of the anti-Stat5b antibodies (top most arrow). Lanes 1, 3, and 7, no prolactin; lanes 2, 4, and 8, prolactin-induced complexes; lanes 5 and 9, prolactin-induced complexes in the presence of 10 µg of anti-Stat5a antiserum; lanes 6 and 10, prolactin-induced nuclear complexes in the presence of 10 µg of anti-Stat5b antiserum.

MMTV transcription is prolactin inducible. Since the complex that bound specifically to the ML-STAT region of the MMTV(C3H) LTR was prolactin inducible, we examined whether viral transcription was affected by this hormone. A normal murine mammary gland cell line transfected with an infectious molecular clone of MMTV, called NMgCl1 (51), was used for this analysis. The cells were grown to confluence, cultured in 1% serum for 48 h, and then induced with prolactin or dexamethasone for 14 to 16 h. Northern blot analysis was performed with a probe specific for the MMTV envelope gene (37). As seen in Fig. 6, there was an approximately two- or fivefold induction of MMTV RNA in cells grown in the presence of prolactin or dexamethasone, respectively. Moreover, prolactin acted synergistically with dexamethasone to further increase transcript levels, since there was a ninefold induction in cells grown with both hormones. Thus, prolactin increased the production of steady-state levels of MMTV RNA, probably through activation of the JAK-STAT pathway.

View larger version (73K): [in this window] [in a new window]   FIG. 6.   Prolactin induction of MMTV RNA. NMuMG cells containing a molecular clone of MMTV (HYB PRO) were grown for 14 h in the presence (+) or absence () of dexamethasone (DEX; 0.5 µM) or prolactin (PRL; 5 µg/ml). Five micrograms of total RNA was subjected to Northern blot analysis and hybridized to an MMTV envelope probe that detects the full-length (upper arrow) and spliced (lower arrow) viral transcripts. The bottom panel shows the same blot hybridized to a mouse -actin probe. Numbers at the bottom represent fold induction after hormone treatment.

MMTV transcription in Stat5b and Stat5a knockout mice. If Stat5b is required for high-level mammary gland expression of MMTV, then mice lacking this transcription factor should have little or no viral RNA in this tissue. RNA was prepared from the mammary glands and lymphoid tissue of mice with targeted mutation of the Stat5b gene, and RNase protection analysis was carried out to examine the level of endogenous MMTV expression. The level of Mtv-17 RNA was dramatically reduced in the Stat5b knockout mice in comparison to a wild-type mouse derived from the same cross (Fig. 7A). Mtv-17 is normally highly expressed in the mammary gland but not lymphoid tissue of inbred mice containing this endogenous provirus. A similar decrease in another endogenous MMTV expressed in mammary gland, Mtv-3, was also seen (not shown). In contrast, the Stat5b knockout mice had normal levels of lymphoid tissue-expressed proviruses Mtv-3 (Fig. 7B) and Mtv-9 (not shown).

View larger version (31K): [in this window] [in a new window]   FIG. 7.   Transcription of endogenous MMTVs in Stat5a and Stat5b knockout mice tissue. RNA isolated from the lactating mammary glands (A) and spleens (B) of Stat5a and Stat5b knockout mice was subjected to RNase T1 protection analysis, using probes specific to the Mtv-17 (A) or Mtv-3 (B) provirus (15, 17). The control mice (+/+) were derived from the same cross that generated the Stat5a and Stat5b knockout animals. The bottom panel shows the same RNAs subjected to Northern blot analysis after hybridization to a mouse -actin probe.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

MMTV is unique among the murine and perhaps other retroviruses in encoding a Sag whose activity it uses as part of its infection pathway. Because this protein causes profound deletion of cognate T cells, any mouse that is infected with an exogenous virus encoding a Sag with the same specificity as its endogenous loci cannot be infected with this virus. As a result, mice with such endogenous viruses do not acquire exogenous MMTV in the mammary gland, nor do they develop mammary tumors (14, 21).

Balanced against the selective pressure to retain endogenous copies of MMTV to protect against exogenous infection is the pressure to retain only those copies of the virus that are not infectious. Inbred strains of mice that retain infectious endogenous proviruses very rapidly succumb to mammary tumors (2, 57). Endogenous MMTVs that are beneficial to mice acquire inactivating mutations that prevent functional virion synthesis or modulate tissue-specific expression of the virus. The only requirement for protection against exogenous virus infection is expression in antigen-presenting cells, such as B cells. Thus, alteration of a transcriptional regulatory region of MMTV so that it is transcribed only in lymphoid cells, as was described here for Mtv-7, prevents it from being a mammary tumor-causing agent and results in protection from infection by exogenous MMTVs.

Identification of the mutation that prevents expression of Mtv-7 and other endogenous proviruses in mammary gland provided us with a tool to look for a transcriptional control element that is required for expression in this tissue. The results presented here show that the T · AG · C change at bp 520 dramatically altered the biology of the virus. All known MMTVs expressed in mammary gland have a T · A base pair at this position, whereas those containing a G · C pair are transcribed only in lymphoid tissue. This T · A base pair is critical for MMTV's transmission as a milk-borne virus. Novel recombinants between Mtv-7, which is not expressed in mammary gland and cannot be transmitted as an exogenous virus, and an exogenous virus, BALB14, acquired the T · A base pair at position 520 yet retained the Mtv-7 Sag specificity. The generation and selection of such recombinants in vivo emphasizes the importance of this T · A residue for mammary gland-specific expression of MMTV.

The single base pair alteration in the Mtv-7 LTR that prevents its expression in mammary gland falls in a sequence that resembles the consensus binding site for the STAT family of transcription factors. We showed here that there is a factor in mammary gland nuclear extracts that preferentially binds to the ML-STAT sequence from MMTV(C3H) but no the L-STAT sequence from Mtv-7. Moreover, Stat5b or a factor that cross-reacts with anti-Stat5b antiserum is part of the complex that recognized this sequence. Both Stat5b and Stat5a have been shown to confer prolactin responsiveness to the -casein (32) and WAP (30) genes, although it is thought that Stat5a is predominantly responsible for the prolactin induction of milk protein gene expression (32). However, in mice that lack the Stat5a gene through targeted mutagenesis, WAP but not -casein or -lactoglobulin RNA levels were greatly diminished (33).

The level of MMTV transcription was reduced in the mammary gland lacking the Stat5a and Stat5b genes. Although the lack of MMTV transcription in the Stat5a knockout mice may indicate a role for this factor, these mice have been shown to have dramatically reduced levels of Stat5b protein (33). Thus, the reduced expression of MMTV may be due to a lack of Stat5b in these mice.

Although MMTV transcripts were diminished in the mammary glands of the Stat5a and 5b knockout mice, they were not affected in lymphoid tissue. This finding argues that although these transcription factors are found and function in lymphocytes, other factors play a dominant role in these cells. Indeed, the T · AG · C change at bp 520 does not affect lymphocyte transcription of MMTV, also indicating that control of expression does not rely on STAT5 factors in these cells. Interestingly, in mice with large copy numbers of the L-STAT-containing transgene, we did detect high levels of both lymphoid and mammary gland tissue expression. This could be due to integration site effects or perhaps because the factors that control transcription through the ML-STAT sequences also weakly interact with the L-STAT sequence; the large number of L-STAT copies present in the same region of the chromosome may alter the kinetics of binding of such factors, thereby allowing transcription. Indeed, the upper complex of the doublet formed on the ML-STAT probe was also weakly found on the L-STAT probe (Fig. 3 and 5).

Stat5b may play an important role in the control of gene expression in the mammary gland that is not solely dependent on prolactin induction. In support of this possibility, tissue culture transfection experiments have shown that Stat5b but not Stat5a could activate gene expression in the absence of this hormone (32). We also found that the complex formed on the ML-STAT oligonucleotide was present in the absence of prolactin, although prolactin both induced the factor(s) bound to this oligonucleotide and caused increased MMTV transcription in a mammary gland tissue culture cell line. Moreover, although MMTV RNA levels increase dramatically in response to lactation in vivo, due at least in part to increased glucocorticoids and progesterone and perhaps to prolactin, MMTV transcription also occurs in nonlactating mammary gland (41). This finding indicates that MMTV expression in the mammary gland is dependent on a transcription factor that is active in the absence of prolactin. Since a number of other growth factors activate Stat5b (as well as Stat5a), it is possible that activation of this factor in the mammary gland occurs predominantly through cytokines other than prolactin.

This is the first example of a STAT consensus sequence affecting a gene's tissue-specific expression, rather than just its ability to be induced by a growth factor or mitogen. Indeed, mutation of the WAP or -lactoglobulin STAT consensus sequence decreased but did not eliminate mammary gland expression in transgenic mice (7, 30). The STAT consensus sequence in all milk protein genes differ from that found in MMTV (see Table 1). For example, there is a C at position 519 in the MMTV sequence instead of the T that is usually seen in the binding site for all STAT factors. Whether this change affects the interaction of Stat5b or some other factor with the MMTV sequence is not yet known. Although none of the WAP or -lactoglobulin constructs used to make transgenic mice had a mutation in an equivalent position in the STAT sequence, substitution of a C for a T in this position in the STAT sequence from the Ly6E/A gene completely abolished binding of STAT1 and STAT3 (28).

How is tissue specificity affected by Stat5b, since this factor is expressed in many tissues and induced by different cytokines (24, 49)? It has been shown recently by a number of groups that STAT factors form complexes with a number of transcription factors, including SP1 (34), RF (20), c-Jun (48), and the GR (53). One possibility is that mammary gland-specific expression is achieved by a complex formed between Stat5b and an as yet undiscovered factor and that this heterologous complex is required to activate MMTV transcription in this tissue. The recognition site for this unknown factor could be near or overlapping that of Stat5b or at some other location within the LTR. For example, we and others have shown previously that sequences between bp 28 and 207 in the MMTV LTR are important for mammary gland-specific expression (9, 27, 40, 41, 47). Alternatively, immediately adjacent to the STAT binding site in MMTV is a region with high homology to the rat -casein gene (Fig. 8). Indeed, 18 of 21 bp are identical between the two genes in this region. It is possible that this region is a recognition site for an as yet unidentified factor that interacts with Stat5b and that this complex controls transcription from MMTV and perhaps the -casein gene.

View larger version (6K): [in this window] [in a new window]   FIG. 8.   Comparison of the sequences adjacent to the rat -casein and MMTV STAT sites. Italics, region of homology between the casein and MMTV sequences; double underline, STAT binding sites; dashes, missing nucleotides; boldface letters, YY1 binding site in the casein gene that prevents STAT5 from binding (39, 45). No YY1 site is seen in the MMTV LTR.

Using transient transfection assays of plasmids bearing the cloned GR and prolactin receptor (PrlR), STAT5, and a -casein reporter construct, Stocklin and colleagues showed that prolactin interfered with glucocorticoid induction (53). Because they also found that the STAT5 and GR proteins complexed with each other, they proposed that the GR sequestered STAT5 and acted as a negative regulator of transcription. We showed here that glucocorticoids and dexamethasone showed synergistic activation of MMTV transcription in a stably transfected cell line. The differences in our results could be due to the level of expression of the GR, PrlR, or STAT5 protein; most likely the endogenous GR, PrlR, and STAT5 levels in NMuMG cells are much lower than those achieved in transient transfection. Similarly, it is possible that these transcription factors interact differently with each other in the context of a stably integrated transcription unit than on an unintegrated plasmid. In support of this possibility, it has recently been suggested that this mechanism of negative regulation, termed squelching, may occur in transient transfection assays but not on genomic DNA (43). Finally, it is also possible that these two transcription factors form a complex that is distinct from that formed on the -casein gene or do not directly interact together at all on the MMTV regulatory sequences.

In the -casein gene, the region thought to control STAT-regulated expression also contains a duplicated STAT element and a YY1 binding site (Fig. 7). It has been shown that this YY1 binding site prevents STAT5 binding and represses -casein gene transcription (39, 45). There is no YY1 consensus sequence adjacent to the STAT binding site in the MMTV LTR, and so there may indeed be different regulation of MMTV and the -casein gene. However, SATB1, which binds to the MMTV LTR at two positions (bp 688 to 717 and 908 to 927), does repress MMTV transcription (31). Since both SATB1 and YY1 are nuclear matrix binding proteins (19, 42), these factors could function similarly in conjunction with STAT factors in the regulation of MMTV and -casein transcription, respectively.

The majority of MMTVs are expressed in lymphoid tissue in addition to mammary gland. It has been reported that cytokines such as interleukin-2 and interleukin-5, as well as lipopolysaccharide, a B-cell mitogen, induce MMTV transcription in B cells(26, 35). These cytokines and lipopolysaccharide induce transcription of a number of genes through activation of the JAK-STAT pathway (13, 23, 25, 55). Interestingly, the casein (18) and WAP (30) genes have been reported to be transcribed in T lymphocytes at low levels. Determination of whether the STAT sequences in the MMTV LTR function to direct expression in lymphoid cells or to mediate response to cytokines awaits analysis similar to that described here for the mammary gland.